Service levels

Service levels: the targets we run to.

These are the operational targets Debt Digest runs to, published so a procurement reviewer can see them before a pilot. They are not a contractual commitment: the service level exhibit to our master services agreement is not yet issued, and no availability percentage is warranted in the agreement until it is. Every number on this page, including response times, support hours and export turnaround, is a target we run to and not a term you can enforce, with one exception named under Incident notification.

Availability

API uptime
99.9% monthly target, not a warranty. Measured against the public /api/health endpoint, computed as a rolling 30-day window. Excludes scheduled maintenance announced ≥ 72 hours in advance.
Maintenance windows
Scheduled outside US-Eastern business hours (8am–6pm ET, Mon–Fri) where possible. Notice: 72 hours minimum, posted on /status and emailed to the creditor contact-of-record.
Geographic redundancy
Single-region (US-East) as of v1.0. Multi-region failover is on the roadmap; will be revisited when the first creditor exceeds 5,000 placed accounts.

Response time (functional SLOs)

Account placement (bulk upload)
Processing target: under 5 minutes per 1,000 accounts. Today placements are held pending manual bankruptcy and deceased-account screening, and generated notices sit in a queue a person releases, so end-to-end time is set by that review, not by processing.
API median latency (p50)
< 200ms measured server-side. Excludes upstream sub-processor latency.
API tail latency (p95)
< 500ms. Status page surfaces "degraded" if exceeded for 5 minutes consecutively.
Settlement offer handling
An offer that meets the creditor’s matrix floor on a matrix cell the creditor has set to accept automatically is recorded as accepted at the moment it is submitted; this is a status decision on the account, not a message to the consumer. Every other offer, including one that meets the floor on a cell the creditor has not set that way, routes to review by the creditor’s own staff, on the creditor’s own clock. No consumer-facing message is ever sent automatically.
Cease-comm enforcement
Same request. A cease request in the consumer portal writes the FDCPA §1692c(c) audit entry and suppresses Debt Digest outbound email and revokes SMS consent in the Debt Digest record across that consumer's accounts, then notifies the creditor of record. Per-account suppression of calls and SMS is set by the participant that receives the request, not by the portal control.

Incident notification

Support

Standard response
Target: 4 business hours for a ticket opened via partner success during a pilot. Support response times become contractual only where an executed order form states them; none is stated today.
NCUA examiner inquiry
We produce an audit-log export (CSV or PDF) for the accounts in scope. We target one business day from request; the export is operator-run and, per the master services agreement, regulator responses go through counsel review, so no delivery time is committed.
Consumer-facing support
Cease-communication request: immediate platform suppression. Dispute filing: immediate in-product acknowledgement; verification follows the responsible participant's evidence workflow. The hold does not expire because time elapsed.

Service credits

None are issued today. The pilot is fee-suspended ($0 platform fee) until 100 accounts are placed or 30 days elapse, whichever comes first, unless the pilot order form states a higher cap or a longer period; the free-period clock does not begin until we confirm in writing that bankruptcy and deceased-account screening is operational. Reaching either limit pauses new placements and never triggers a charge, so there is no fee to credit. Any credit schedule for paid subscriptions will be set in the service level exhibit when it is issued, and will take effect only by amendment to a signed agreement.

What these targets do not cover

Verification: uptime is reported on /status from our own first-party monitoring. No independent third-party monitor is engaged yet, so there is no external evidentiary record of uptime.